Occupational Epidemiology

1
Occupational Epidemiology

• ??. ????? ?????????????
• ??????????????????
• ?????????????????

2
Occupational epidemiology study of the
frequency and the causes of work-related diseases
and injuries

• Branch of epidemiology that is defined by the
  exposure rather than the outcome.
• Helps in studying exposures and outcomes that are
  rare in the general population.
• Helps in devising occupational exposure
  guidelines.
• Helps in deciding what remedial measures to
  recommend.

3
Examples of the development of occupational
epidemiology
Time London chimney sweeps Miners
Late 1700s Late 1800s 1920-1940 1920-1940 Excess of scrotal cancer found among chimney sweeps in London. Carcinogenicity of coal tar products recognized in different industries. Experimental model of soot carcinogenesis demonstrated. Excess of pneumonia found among gold and silver miners in Germany. Excess of respiratory cancer found among other underground metal miners. Survey finds that 50 of miners deaths are due to lung cancer, 25 due to non-malignant respiratory disease. Causative agent for lung cancer in miners found ionizing radiation from uranium and radium deposits in the mines
4
Identifying Occupational Hazards

• Recognize a disease cluster among workers from
  particular occupations or industries.
• Conduct a survey in the industry to determine the
  magnitude of the problem.
• Consider other diseases which may occur at
  excess.
• Determine exposure to a known hazardous agent or
  to another agent not yet known to be hazardous.
• Or
• Start out with a particular exposure and conduct
  medical surveillance.

5
Characterizing the workplace environment(exposure
assessment)

• Identify agents likely to be toxic.
• This may be easy (e.g. asbestos exposure) or very
  difficult (i.e. mixtures of chemicals).
• Can use information from prior research or
  consult toxicologists.
• Can try to gain information by determining the
  part of the manufacturing process that seems be
  most hazardous, or by looking at the type of
  illness caused by the unknown agent.
• Establish the most relevant routes of exposure
  for the agents of concern.
• Measure the exposure.

6
Measuring Exposure

• Type of exposure data
• Quantified personal measurements
• Quantified area/job specific data
• Ordinal ranked jobs/tasks
• Duration of employment in the industry at large
• Ever/never employed in the industry

best
worst
7
Question Should we measure

• Point exposure
• Cumulative exposure
• Highest exposure
• Length of exposure
• This depends on a variety of issues including the
  particular agent and the availability data.

8
Collecting exposure data

• Direct measurement over a period of time
• Problems
• Important exposures in the past are missed.
• May lead to overestimation when the exposure is
  only measured in areas where it is assumed to be
  highest.
• May lead to underestimation when the measuring
  device is placed away from the work area in order
  not to interrupt the work.
• Wearing a measuring device may alter a persons
  behavior.

9

• Compiling an inventory of existing data and
  determining which data are most complete and
  useable for the study.
• a. Historical exposure reconstruction
• b. Concurrent and prospective exposure estimation
• Problems a. - Past data may not exist or may be
  incomplete.
• - It may not be possible to combine
  data from different time periods.
• (different measurement techniques may have
  been used similar job categories may have
  meant different exposures.)
• b. No useable data may exist
• For both, historical exposure reconstruction and
  concurrent and prospective exposure estimation,
  we can get information from

10

• Industrial hygiene data
• (May lead to overestimation when the exposure is
  only measured in areas where it is assumed to be
  highest may lead to underestimation when the
  measuring device is placed away from the work
  area in order not to interrupt the work
• wearing a measuring device may alter a persons
  behavior)
• Process descriptions
• (can be used to identify and localize potential
  agents)
• Plant production records
• (can be used to determine the introduction and
  removal of chemicals and to detect seasonal
  variations)

11

• Inspection/accident reports
• (can be used to detect unusual exposures and to
  distinguish between routine and excess exposure)
• Engineering control/protective equipment
  documentation
• (can be used to determine if the workers were
  fully exposed or if they were protected)
• Biological monitoring results
• (e.g. blood, urine,monitoring the usefulness
  depends on the agent)
• Could use a scheme such as high, moderate, low,
  possible, no exposure and use information from
  personnel records (e.g. job title, pay code,
  dates of employment,) to assign workers to the
  different groups
• (Assigning jobs/work areas to different exposure
  levels is difficult misclassification is
  likely.)

12

• For concurrent and prospective exposure
  estimation we get additional information by
• Updating personnel files
• Collecting additional exposure data
• (toss-up between measuring and recording
  everything and starting to monitor only when a
  significant health hazard is noticed it is
  sometimes suggested to routinely take a sample of
  measurements)
• Conducting ecologic studies (compare disease
  rates and industrial activities between different
  areas)
• Problem is the ecological fallacy

13
Combining exposure data from various sources

• Times
• Work areas
• Industries
• Countries

14
Purist approach

• Only workers with the most detailed measurement
  values can be used.
• Problem
• Measurement error is reduced and validity is
  increased, but sample size and thus precision and
  drastically reduced.

15
Take everybody approach

• Take everybody who has a minimum of useable
  information.
• Problem
• Difficult to combine site/times with measured
  concentrations and sites/times with nothing but
  job information.
• Job classifications may differ between different
  times, industries, or countries.

16
Study Designs

• Case series
• Identification and reporting of a disease
  cluster. The cluster might be found among the
  work force as a whole or among some segment of
  the work force.
• Case series can be very useful to start an
  epi-investigation, especially when the disease is
  extremely rare and the causal factors are
  unknown.
• Disease clusters can be misleading, however,
  since they could be entirely due to chance.

17
Study Designs

• Cohort Studies
• Most accepted study design since it most closely
  resembles the experimental setting (exposure ?
  disease).
• The study includes the entire available and
  disease-free study population.
• 2.1 Prospective cohort study
• 2.2 Historical cohort study
• 2.3 Sub-cohort analyses

18
2.1 Prospective cohort study

• The cohort is enumerated at the time of the
  study, cohort members are followed into the
  future.
• The rates of disease occurrence are usually
  compared to the rates in the national or regional
  population to determine which diseases occur more
  or less frequently among the workers. SMRs or
  SIRs are calculated.
• Prospective cohort studies are rarely used, since
  they take too long, are too expensive, and are
  not appropriate for rare diseases. However, they
  are appropriate for consequences of an
  occupational exposure that occur within a brief
  time span (approximately 5 years or less).
• They are also useful in medical surveillance,
  where the cohort is followed into the future and
  the workers health status and the occurrence of
  disease in the cohort is determined. The focus of
  a medical surveillance program may be very narrow
  or wide.

19
2.2 Historical cohort study

• Past records are used to enumerate the cohort.
  The cohort is then followed into the present.
• Historical cohort studies are cheaper and take
  less time than prospective cohort studies SMRs
  and SIRs can be calculated.
• However, records on the outcome may not be
  available. Thus, historical cohort studies are
  mostly used for fatal diseases so that death
  certificates can be used to determine the type of
  illness and the time of death.
• Data on non-fatal diseases are only available
  when special efforts have been made to collect
  them (e.g. cancer registries).

20
2.3 Sub-cohort analyses

• Comparisons are made between subgroups (e.g.
  high/medium/low exposure) rather than between the
  workers and the general population.
• Sub-cohort analyses can be conducted in
  prospective or historical cohort studies, but
  direct age adjustment must be used and SRRs
  (standardized rate ratios) must be calculated.
• SRR expected cases in the reference population
  based on the rates in the exposed group
• observed cases in the reference population
• Since sub-cohort analyses are expensive they are
  generally only performed for diseases with an
  overall mortality or morbidity only performed for
  diseases with an overall mortality or morbidity
  excess and for diseases of special interest.

21
3. Case-control studies

• Smaller sample size, shorter time frame, and thus
  reduced cost. ORs are calculated.
• 3.1 Nested case-control study
• 3.2 Registry based case-control study

22
3.2 Nested case-control study

• A nested case-control study is a case-control
  study embedded in a cohort study.
• It is useful for workplace hazards of particular
  interest that cannot be studied efficiently with
  a cohort or sub-cohort analysis.
• Example Solvents ? Leukemia
• It would be a huge task to reconstruct the
  exposures of a large cohort of workers over a
  long period of time. Instead leukemia cases are
  identified during follow-up and are used as the
  cases. Leukemia free workers are used as
  controls.

23
3.2 Nested case-control study

• Sometimes an occupational cohort cannot be
  enumerated (e.g. farmers, auto mechanics).
• In this case a registry can be used to define
  cases and controls. (E.g. cancer registry,
  hospital admissions, insurance claims, disability
  pension awards,)
• The cases can be taken from the registry.
• The controls can be taken from registrants with
  other diseases or from the source population for
  the registry.

24
3.1 Nested case-control study

• Most registry based case-control studies lack
  detailed exposure data. Often only the type of
  industry or the job title are known. Therefore,
  since they are less informative than a nested
  case-control studies, registry based case-control
  studies are mostly used for screening hypotheses.

25
4. Proportionate mortality studies

• A proportionate mortality study is conducted when
  information on occurrence of disease or death
  exists, but it is impossible to enumerate the
  cohort.
• Ex. Death certificates are available, but
  personnel information is unavailable or
  incomplete.
• We can compare the proportional distributions of
  causes of death among the workers with the
  corresponding proportions in the reference
  population.
• This gives us an indication of the relative
  disease frequency.

26
4. Proportionate mortality studies

• Advantage Quick and inexpensive
• Disadvantage The identified deaths may not be
  representative of all deaths that would have been
  identified had the cohort been enumerated and
  followed.
• Example Sick people may have causes of death
  must add up to 100. Therefore, an excess of
  deaths from one cause necessarily leads to a
  deficiency of deaths from one or more other
  causes. Thus, a deficiency of deaths from one
  cause of death does not imply that the exposure
  is protective against this disease.

27
5. Cross-sectional studies

• Disadvantage Retirees, transferred workers,
  laid-off workers, dead workers and workers who
  quit for health reasons are missed. Thus the
  potentially most important workers are missed.

28
Study Validity

• Selection bias
• Ex.
• Higher response rate among the most heavily
  exposed people with the disease.
• Healthy worker effect (healthy workers are more
  likely to gain and remain in employment).
• Note As the cohort is followed over time the
  effect of the healthy worker effect on the study
  results decrease
• Note The healthy worker effect can be minimized
  by choosing other active workers rather than the
  general population as the comparison group.

29
Information bias

• Non-differential The likelihood of
  misclassification is the same for the compared
  groups.
• Ex.
• The study outcome is not well defined and
  includes a wide range of etiologically unrelated
  outcomes. This may obscure the effect of the
  exposure on one specific outcome (a large
  increase in this outcome may only produce a small
  increase in the overall group of outcomes
  studied).
• The exposure of interest is not well defined
  (i.e. an exposure occurring shortly before the
  diagnosis may be incorrectly included).
• This bias is of particular concern in studies
  that show no association between the exposure and
  the outcome.

30

• Differential The likelihood of misclassification
  of the exposure is different for the diseased and
  the non-diseased.
• The likelihood of misclassification of the
  disease is different for the exposed and the
  non-exposed.
• Note
• It is sometimes worth decreasing the sample size
  (and thus increasing random error) if the
  increased accuracy we can achieve on fewer study
  subjects greatly decreases misclassification due
  to information bias.

31
Recall bias

• Note studies have been performed to determine
  how well current workers recall their work
  history (Baumgarten et al., 1983 Brisson et al.,
  1988).
• They found that approximately 80 of the person
  years were correctly identified (identification
  was more accurate for the past 12 years and less
  accurate for years lying further back). Recall
  did depend on the number of jobs workers held
  (the more jobs the less accurate the recall), but
  did not depend on age or level of education.

32
Confounding

• The following confounders are often considered in
  occupational studies
• Gender
• Ethnicity
• Smoking
• SES
• Time related factors

33
Time related factors

• Length of follow-up (time of hire until disease
  onset, death, or end of study)
• Duration of employment (time of hire until
  termination of employment strongly associated
  with cumulative exposure)
• Age at hire
• Age at risk (age at any point during follow-up)
• Calendar year
• These factors are associated with the outcome
  either directly or through their influence on the
  healthy worker effect.

34
Time related factors

• Length of follow-up (time of hire until disease
  onset, death, or end of study)
• Duration of employment (time of hire until
  termination of employment strongly associated
  with cumulative exposure)
• Age at hire
• Age at risk (age at any point during follow-up)
• Calendar year
• These factors are associated with the outcome
  either directly or through their influence on the
  healthy worker effect.

35
Examples

• The older the workers the more likely they are to
  get ill or to die. Thus age at risk is
  associated with the outcome.
• Disease incidence may change over time. Thus
  calendar year may be associated with the outcome.
• The healthy worker effect is most pronounced
  immediately after the workers are hired (i.e.
  when they are healthy enough to be employed).
  Around 15 years after they were hired the healthy
  worker effect almost disappears. Thus, length of
  follow-up influences the healthy worker effect
  and therefore the outcome.

36

• Mortality is lowest (and thus the healthy worker
  effect is strongest) among those with the longest
  duration of employment. Thus, duration of
  employment influences the healthy worker effect
  and therefore the outcome.
• The healthy worker effect is stronger among
  workers hired at an older age than among young
  workers. Thus, age at hire influences the
  healthy worker effect and therefore the outcome.

37

• If the time related factors are also associated
  with the exposure they act as confounders.
• Examples
• Older workers may have been exposed to different
  chemicals in the past.
• Workers hired at an older age may be assigned to
  different jobs.
• Workers with a long duration of employment may
  have different jobs.

38
Reference

• Annette Bachand, Introduction to Epidemiology
  Colorado State University, Department of
  Environmental Health
• Leslie Gross Portney and Mary P. Watkins (2000).
  Foundations of Clinical Research Applications to
  Practice. Prentice-Hall, Inc. New Jersey, USA